Abstract: Abstract: Drought risk management is an effective countermeasure to drought disaster mitigation. Process identification and frequency analysis of drought characteristics are the basis for risk management. Currently, the major issues associated with drought process identification are: 1) consistency of hydrological drought indicators (e.g., surface runoff, soil moisture content, groundwater depth) is susceptible to the impact of human activities; 2) definitions of threshold values in drought process identification are usually lack of clear physical meaning, and the beginning and end time of drought process vary with these values; 3) traditional drought duration defined as the number of periods from the beginning to end time of drought process, may lead to identify the drought event with low intensity and several drought alleviation periods as the severe one due to its long duration; 4) drought intensity lacks time comparability. Therefore, in this study, we used the Palmer drought severity index (PDSI) method by taking the meteorological factors of less influenced as the input to determine the dry/wet states and the PDSI value at each period. Determination of the dry/wet states and the PDSI value at each period included five main steps: hydrological accounting based on meteorological and soil water characteristics data, calculation of climatic coefficients, analysis of the amount of climatically appropriate for existing conditions (CAFEC) precipitation, calculation of Palmer moisture anomaly z-index based on actual and CAFEC precipitation, and estimation of the dry/wet states and the PDSI value based on z-index. The dry/wet states consisted of dry period, wet period, transition period, transition period in dry spell, and transition period in wet spell. Then the drought process identification criteria were analyzed based on the dry/wet states and the PDSI value by trial and error method according to actual drought in the study area. Drought duration was defined as the number of periods under dry states, while the drought severity was estimated as summation of absolute value of negative moisture anomaly index, which was of time-comparability during a drought event. On the basis of the distribution of drought duration and drought severity generated by the frequency curve fitting method, their joint distribution was constructed via the GH Copula, and accordingly the estimation of drought recurrence periods at Kunming station from 1951 to 2011 were conducted. There were 43 droughts were identified, and the drought that occurred in 2009-2010 was the most severe drought event in the study period with the recurrence period of 64.7 years. Results showed that drought process indentified based on the PSDI model was consistent with the actual regional drought circumstances and the results by using runoff index methods, notwithstanding the disparity from that by applying the precipitation index that were not concerned with surface hydrological process. The proposed drought frequency analysis was of clear physical concept, and the expression of drought duration and severity was of physically reasonability. The results of this study can provide a reference for assessment of drought disaster hazard.